Editor's Note: For a long time, Fintech and DeFi were seen as two unrelated financial systems: one compliant, centralized, and valuable; the other open, on-chain, and more like public infrastructure. Fintech excels at turning traffic into revenue, while DeFi excels at maximizing efficiency. With tokenization, stablecoins, and on-chain transactions increasingly penetrating traditional systems, integration is irreversible. So, will the future of finance involve building tollbooths on an open path, or will the tollbooths eventually learn to embrace openness?
This article compares and analyzes Fintech and Onchain Protocols from four dimensions: revenue, scale, number of users, and commission rate, attempting to answer whether DeFi or Fintech will win.
The following is the original text:
Gm, Fintech architects, today we're bringing you some truly groundbreaking content.
We have partnered with analytics firm Artemis (often referred to as the "Bloomberg Terminal of Digital Finance") to officially release the first-ever comparative analysis of Fintech and DeFi key performance indicators (KPIs).
If you've ever struggled with whether Robinhood or Uniswap is a better investment, then you've come to the right place.
summary
We put fintech stocks and crypto tokens on the same "comparison table," a true side-by-side comparison.
Covering multiple sectors including payments, digital banking, transactions, lending, and prediction markets, we compared revenue, user scale, take rate, key industry metrics, and valuation metrics.
The result was quite unexpected:
Hyperliquid's trading volume has reached more than 50% of Robinhood's.
The outstanding loan balance of the DeFi protocol Aave has surpassed that of the buy-now-pay-later platform Klarna.
The growth rate of stablecoin payment channels is far faster than that of traditional payment service providers.
The user base of wallets like Phantom and MetaMask is now comparable to that of newer banks like Nubank and Revolut.
We found that valuations truly reflect this tension: crypto assets are either heavily discounted due to the uncertainty of "how to monetize them in the future," or given an extreme premium due to overly high expectations.
Ultimately, we raise a core question about "convergence": Will the crypto world learn to build tollbooths, or will fintech ultimately adopt crypto's open payment and clearing channels (open rails)?
Objects involved
Block, PayPal, Adyen, Tron, Solana, Coinbase, Robinhood, Uniswap, Aave, Affirm, Klarna, Polymarket, DraftKings
Two financial systems
For years, we have viewed the crypto world and fintech as two parallel universes: one regulated, audited, and traded on NASDAQ, and the other a permissionless system where assets circulate on both decentralized and centralized exchanges.
They all use the same language—revenue, transaction volume, payments, lending, and trading—yet they carry drastically different "accents." And this landscape is changing.
With Stripe's acquisition of Bridge, Robinhood's launch of prediction markets, and PayPal's issuance of its own dollar stablecoin, the boundaries between the two systems are becoming increasingly blurred.
The real question is: when these two worlds actually meet and collide, how should they be compared on the same comparison table?
So we decided to conduct an experiment.
We selected fintech companies you are familiar with—covering payment processors, neobanks, buy-now-pay-later (BNPL) institutions, and retail brokers—and compared them one by one with their respective crypto-native counterparts.
In the same set of charts, we use familiar fintech metrics for comparison (such as P/S, ARPU, TPV, number of users, etc.): green bars represent stocks, and purple bars represent tokens.
Thus, a picture of two financial systems gradually emerges: on-chain financial protocols often match or even surpass their fintech counterparts in terms of transaction volume and asset size; however, the economic value they capture is only a small fraction of it; compared to comparable fintech companies, crypto assets are priced either at an extreme premium or a deep discount, with almost no "middle ground"; and in terms of growth rate, the gap between the two is even more significant.
Payment: The "pipeline" for the flow of funds
Let's start with the largest sector in fintech – transferring money from point A to point B.
On the green side, there are a group of true giants:
PayPal: Processes $1.76 trillion in payments annually
Adyen: Handling $1.5 trillion
Fiserv: This "infrastructure company nobody mentions at the party" handles $320 billion
Block (formerly Square): Facilitates the flow of $255 billion through Cash App and its merchant network.
These companies constitute the most mature and stable "funding pipeline layer" in the traditional fintech system.
On the purple side, here is the annualized B2B payment volume estimated by the Artemis stablecoin team:
Tron: Stablecoin transfers totaled $68 billion
Ethereum: $41.2 billion
BNB Chain: $18.6 billion
Solana: Approximately $6.5 billion
In absolute terms, the two are not on the same scale. The total stablecoin transaction volume across all major public blockchains is only about 2% of the size of fintech payment processors. If you squint at that market share chart, the purple bars are almost just a rounding error.
But what's really interesting is the growth rate.
PayPal's total payment volume only grew by 6% last year.
Block growth 8%
Adyen, a European darling, achieved 43% growth—quite impressive by fintech standards.
Let's take a look at blockchain:
Tron: Growth of 493%
Ethereum: Growth of 652%
BNB Chain: Growth of 648%
Solana: 755% year-on-year growth, the fastest growth rate.
To reiterate, these figures are from B2B payment estimates compiled by the Artemis data team based on McKinsey research.
The results are very clear: the growth rate of stablecoin "rails" is far faster than that of traditional fintech payment systems.
Of course, their starting point is also much smaller.
The question then becomes: Who is truly capturing economic value?
Fiserv: It takes a 3.16% cut from every dollar it handles.
Block: 2.62% extracted
PayPal: 1.68% commission
Adyen: Due to the adoption of a lower-margin enterprise-level model, only 15 basis points (0.15%) are extracted.
These are all real commercial companies whose revenues are directly and stably tied to the scale of payments.
As for blockchain technology, its commission rates for stablecoin transfers and broader asset transfers are much lower, generally between 1 and 9 basis points (bps). Tron covers stablecoin transfer costs by charging users TRX, while Ethereum, BNB Chain, and Solana charge end users gas fees or priority fees.
These public blockchains are very strong at facilitating transfers and promoting asset liquidity, but they take a much smaller percentage of the transaction compared to traditional payment service providers. Avoiding fees such as exchange fees and merchant charges is one of the key reasons why blockchain can claim a significant efficiency advantage over existing payment systems.
Of course, there are also a large number of onchain payment orchestrators who are happy to add more fees on top of the base fees—which presents a great opportunity for “facilitators” to build economic value on top of it.
Neobanks: Wallets become new bank accounts
In the digital banking sector of fintech, there are genuine banks (or banks that "rent licenses"), such as Revolut, Nubank, SoFi, Chime, and Wise. These institutions possess licenses, deposit insurance, and compliance departments.
In the crypto world, we see wallets and yield protocols such as MetaMask, Phantom, Ethena, and EtherFi. These are certainly not "banks," but millions of people store their assets here; and increasingly, people are earning returns on their savings through these platforms. Even with different regulatory frameworks, this functional comparison remains valid.
Let's start with user scale: Nubank has 93.5 million monthly active users (MAU), making it the world's largest digital bank. This scale is built on the backdrop of high smartphone penetration in Brazil and the extremely complex local banking system; Revolut has 70 million users in Europe and other regions; next is MetaMask, with about 30 million users—a scale that has surpassed Wise, SoFi, and Chime.
It's worth noting that most MetaMask users don't use their crypto wallets to pay rent; instead, they use them to interact with DEXs or participate in lending protocols. Another leading wallet, Phantom, boasts 16 million monthly active users. Initially positioned as the best-performing wallet within the Solana ecosystem, Phantom quickly expanded to multiple public chains and now offers its own stablecoin $CASH, debit cards, tokenized stocks, and prediction markets.
Next, let's see where the money is kept.
Revolut: Customer balance of $40.8 billion
Nubank: $38.8 billion
SoFi: $32.9 billion
These are all real deposits, generating net interest income for the institutions.
A highly similar correspondence also exists in the crypto world:
Ethena: This synthetic dollar agreement, which didn't exist two years ago, now holds $7.9 billion.
EtherFi: Scale Reaches $9.9 Billion
These are not "deposits," but rather pledged assets, yield-bearing positions, or liquidity held in smart contracts—known in industry terminology as TVL (Total Value Locked).
From the user's perspective, the logic is not fundamentally different: the money is placed somewhere and continues to generate returns.
The real difference lies in how these platforms monetize their "existing funds" and how much money they can make from users.
SoFi: Average annual income of $264 per user. This is not surprising—SoFi engages in strong cross-selling between loans, investment accounts, and credit cards, and its user base has a higher overall income.
Chime: $227 per person, income mainly comes from exchange fees.
Nubank operates in Brazil, a market with a low GDP per capita of $151.
Revolut: Despite its large user base, the average cost per user is only $60.
What about EtherFi? It costs an average of $256 per person, almost the same as SoFi.
A disadvantage for this crypto upstart is that EtherFi has only about 20,000 active users, while SoFi has 12.6 million.
In other words, this DeFi protocol, with a very small user base, has achieved the same monetization efficiency as top-tier digital banks with a massive user base.
From another perspective, MetaMask generated approximately $85 million in revenue last year, which translates to an ARPU of about $3, even lower than Revolut's early-stage levels.
Although Ethena has a TVL of $7.9 billion, its user reach is still only a fraction of that of Nubank.
Valuation is a direct reflection of this dual tension between "growth" and "monetization ability".
Revolut is valued at approximately 18 times revenue, a pricing that reflects its market positioning and the "option value" of future expansion; EtherFi is valued at approximately 13 times; and Ethena at approximately 6.3 times, roughly in line with SoFi and Wise.
A rather counterintuitive conclusion is emerging: in terms of valuation, the market is treating DeFi/on-chain "banks" and traditional fintech banks in a very similar way.
The so-called "convergence thesis" refers to the idea that wallets will eventually evolve into digital banks.
We've already seen this trend materialize: MetaMask launched a debit card, and Phantom integrated fiat currency deposit and withdrawal channels. The direction is clear; it's just still in progress.
However, when on-chain "digital banks" like EtherFi have higher per capita income than Revolut, the gap between the two is not as large as the narrative suggests.
Trading: On-chain DEXs are approaching the level of traditional brokerages
Let's shift our focus to the capital market.
What really surprised us was that the trading volume of on-chain exchanges is now comparable to that of traditional securities firms.
Robinhood processed $4.6 trillion in transactions over the past 12 months, primarily in stocks, options, and crypto assets, representing an asset size of approximately $300 billion (fluctuating).
Hyperliquid's spot and perpetual contract trading volume is approximately $2.6 trillion, primarily driven by crypto trading, but stocks and commodities are beginning to gain a share.
Coinbase has a trading volume of approximately $1.4 trillion, almost entirely from crypto assets.
Charles Schwab, as an "old-school force," did not disclose its transaction volume in the same way, but its $11.6 trillion in assets under custody is enough to illustrate the size gap between new and old funds—approximately 40 times that of Robinhood.
This also paints a clear contrast: on-chain transactions have approached those of mainstream securities firms in terms of "traffic," but in terms of "existing assets," the traditional system still holds an overwhelming advantage.
Other decentralized exchanges are also not to be overlooked. For example, Uniswap, a protocol that has proven the viability of Automated Market Makers (AMMs), has a trading volume of nearly $1 trillion; Raydium (the leading DEX in the Solana ecosystem) has completed $895 billion; and Meteora and Aerodrome together contributed approximately $435 billion.
In total, the processing scale of major DEXs is now comparable to that of Coinbase. Three years ago, this was almost unimaginable.
Of course, we don't know how much of this trading volume is inflated and how much is genuine trading; but the trend itself is key. Furthermore, while the "convergence" in trading volume is real, DEXs and traditional brokerages have fundamentally different take rates.
Traditional brokerages / Centralized platforms:
Robinhood: Takes 1.06% of each transaction, primarily from Payments for Flow (PFOF) and crypto spreads.
Coinbase: Approximately 1.03%; high spot transaction fees are still quite common among centralized exchanges.
eToro: Even so, it still requires 41 basis points.
DEX, on the other hand, operates in a completely different universe:
Hyperliquid: 3 base points
Uniswap: 9 basis points
Aerodrome: 9 base points
Raydium: 5 base points
Meteora: 31 basis points (a notable exception)
Decentralized exchanges can generate extremely high trading volumes, but their commission rates are significantly suppressed due to fierce competition among liquidity providers (LPs) and traders.
This is quite similar to the division of labor in traditional markets: real exchanges (such as NASDAQ and Intercontinental Exchange) and brokerages that bring clients to trading venues have different functions.
This is precisely the paradox of DEX.
DEXs have built trading infrastructure that can compete head-on with centralized exchanges in terms of trading volume: operating 24/7 with virtually no downtime, no KYC required, and anyone can list coins. However, with a trading volume of $1 trillion, even with a 9 basis point commission, Uniswap would only generate about $900 million in fees and about $29 million in revenue; while with a trading volume of $1.4 trillion, a 1% commission rate would generate $14 billion in revenue for Coinbase.
This difference is accurately reflected in market valuations:
Coinbase: 7.1 times sales
Robinhood: 21.3x (High for brokerages, but supported by growth)
Charles Schwab: 8.0 × (maturity multiple of mature business)
Uniswap: 5.0x Fees
Aerodrome: 4.8 × handling fee
Raydium: 1.3x commission
The conclusion is not complicated: the market did not price these agreements as "high-growth tech companies," and one important reason is that they have lower commission rates compared to traditional brokerages, and therefore less economic value to capture.
The direction of market sentiment is clear from the stock price performance.
Robinhood has risen approximately 5.7 times since the end of 2024, benefiting from a recovery in retail investment and a rebound in the crypto market; Coinbase has risen approximately 20% during the same period; while Uniswap, the protocol that "spawned thousands of DEX forks," has seen its stock price (token) fall by 40%.
Despite the large volume of trading that continues to flow through these DEXs, the associated tokens have not captured a corresponding level of value, partly because their "use" as investment vehicles is not clearly defined.
The only exception is Hyperliquid. Due to its explosive growth in size, Hyperliquid's performance was almost in sync with Robinhood, achieving similar growth rates during the same period.
While DEXs have historically struggled to capture value and have been perceived as a public good, projects like Uniswap have begun to activate their "fee switch"—using transaction fees to buy back and burn UNI tokens. Currently, Uniswap's annualized revenue is approximately $32 million.
We remain optimistic about the future: as more and more trading volume migrates to the blockchain, value is expected to gradually flow back to the DEX tokens themselves, with Hyperliquid being a successful example.
However, for now, until token holders have a clear and direct value capture mechanism like Hyperliquid, DEX tokens will continue to lag behind centralized exchange (CEX) stocks.
Lending: Underwriting the Next Generation of Financial Systems
In the lending process, the contrast becomes even more intriguing.
On one hand, there's the core business of fintech—unsecured consumer credit:
Affirm: Let you split a Peloton bicycle into four installments.
Klarna: Offering the same installment payment options for fast fashion.
LendingClub: It pioneered the P2P lending model and later transformed into a true bank.
Funding Circle: Underwriting loans for SMEs
These companies share a highly consistent profit-making logic: charge borrowers higher interest rates than they pay depositors, and pray that default rates won't eat up this interest rate spread.
On the other hand, there are collateralized DeFi lending platforms: Aave, Morpho, and Euler.
Here, borrowers collateralize ETH, lend USDC, and pay an algorithmically determined interest rate; once the collateral price falls to a dangerous level, the protocol automatically liquidates—there are no collection calls and no bad debt write-offs.
These are essentially two completely different business models, they just happen to both be called "lending".
Let's start with the loan size.
Aave's outstanding loans total $22.6 billion.
This exceeds the sum of the following companies:
Klarna: $10.1 billion
Affirm: $7.2 billion
Funding Circle: $2.8 billion
LendingClub: $2.6 billion
The largest DeFi lending protocol has surpassed the largest BNPL platform in terms of loan volume.
Please stop and truly consider this fact.
Morpho added another $3.7 billion. Euler, which was relaunched after a vulnerability attack in 2023, is currently valued at $861 million.
Overall, the DeFi lending system has grown to a size that rivals the entire listed digital lending industry in about four years—but its economic structure is inverted.
In the traditional fintech sector: Funding Circle's net interest margin is 9.35% (which is related to its business model, which is closer to private lending); LendingClub's is 6.18%; and Affirm, although a BNPL company rather than a traditional lending institution, still achieves 5.25%.
These are all quite lucrative interest rate spreads—essentially compensation for them bearing credit risk and personally conducting credit granting and risk control (underwriting).
In the crypto world: Aave's net interest margin is only 0.98%; Morpho's is 1.51%; and Euler's is 1.30%.
Overall, even with larger loan volumes, DeFi protocols generally earn lower interest spreads than fintech lenders.
DeFi lending is designed to be overcollateralized.
Borrowing $100 on Aave typically requires collateral of $150 or more. The agreement itself does not bear credit risk, but rather liquidation risk—a completely different kind of risk.
The fees paid by borrowers are essentially paying for leverage and liquidity, rather than paying for access to credit that they would not otherwise have been able to obtain.
Fintech lenders, on the other hand, offer unsecured credit to consumers to meet the "buy now, pay later" demand; the interest rate spread exists to compensate those who will never repay.
This will be directly reflected in the loss data caused by actual defaults, and how to manage these default risks is the core work of credit granting and risk control (underwriting).
So, which model is better? The answer depends on what you want to optimize.
Fintech lending serves borrowers who don't currently have money but want to consume now, thus bearing real credit and default risks. It's a brutal business. Early digital lending institutions (such as OnDeck, LendingClub, and Prosper) have repeatedly teetered on the brink of collapse.
Even though Affirm's business is performing well, its stock price has fallen by about 60% from its historical high. This is often because the market uses the valuation logic of SaaS to price its credit revenue, without fully factoring in the inevitable credit losses in the future.
DeFi lending is essentially a leveraged business.
It doesn't serve "people who don't have money," but rather users who already own assets but don't want to sell and just want liquidity; it's more like a margin account. There are no traditional credit decisions here; the only criterion is the quality of the collateral.
This model is highly capital-efficient and scalable, earning very thin interest spreads on a massive scale; however, it also has clear limitations—it is only useful for those who are already on-chain, have a large amount of assets, and want to obtain returns or additional leverage without selling their assets.
Predicting the market: Nobody can say for sure?
Finally, let's look at prediction markets.
This is the latest and most peculiar battleground between Fintech and DeFi. For decades, they have been considered "oddballs" in an academic sense: economists love them, but regulators avoid them like the plague.
Iowa Electronic Markets once ran election prediction services on a small scale; Intrade experienced a brief period of prosperity before being shut down; and many other projects were directly classified as gambling or sports betting.
The idea that "trading the outcomes of real-world events, and that these markets can make better predictions than polls or commentators" has long remained a theoretical concept.
All of this changed in 2024 and accelerated during Trump's second term: Polymarket processed over $1 billion in election betting; Kalshi won its lawsuit against the CFTC and began rolling out political contracts to US users; Robinhood, as always, didn't want to miss any trends and quickly launched event contracts; and DraftKings, which already operated a de facto prediction market through Daily Fantasy Sports, stood by with a market capitalization of $15.7 billion and annual revenue of $5.5 billion.
Prediction markets have finally moved from fringe experiments to the center stage of the financial and crypto world.
This sector has rapidly transitioned from a niche experiment to the mainstream in about 18 months—the weekly trading volume of prediction markets has reached approximately $7 billion, setting a new record.
DraftKings traded $51.7 billion in the past 12 months; Polymarket reached $24.6 billion, about half the size of DraftKings, and it is still a crypto-native protocol, which is theoretically not allowed for use by US users; Kalshi, as a compliant US-based alternative, traded $9.1 billion.
Judging from trading volume alone, Polymarket is already quite competitive. It has built a highly liquid, globally-reaching prediction market on Polygon, while Kalshi is still embroiled in legal battles for compliance.
But when we turn to income, the comparison begins to fall apart.
DraftKings generated $5.46 billion in revenue last year; Kalshi only generated $264 million; and Polymarket only achieved an annualized revenue run rate of approximately $38 million after enabling taker fees for 15-minute crypto markets.
This once again reveals a familiar divide: in terms of "scale," DeFi has caught up; but in terms of "monetization capabilities," traditional financial and gaming companies still hold an overwhelming advantage.
The core difference lies in the take rate – also known as the "hold" in the context of sports betting.
DraftKings: Takes 10.57% of every dollar wagered. This is a typical sports betting model—the bookmaker sets the price, provides the odds, and manages the risk, taking a substantial cut.
Kalshi: The commission is 2.91%, which is lower and more in line with its positioning as a financialized exchange model.
Polymarket: Only 0.15%. With $24.6 billion in trading volume, the revenue that can be captured at present is very limited.
The conclusion is straightforward: the key to predicting market differentiation lies not in "scale," but in "revenue structure."
This is almost a replay of the DEX logic.
Polymarket doesn't focus on value capture; instead, it concentrates on providing infrastructure for prediction markets: matching buyers and sellers and settling contracts on-chain. It doesn't employ odds setters, manage balance sheets, or stand on your side in a bet. The efficiency is indeed remarkable, but monetization isn't its core objective at present.
However, investors clearly believe that Polymarket will eventually become a cash cow. Polymarket is valued at approximately $9 billion, corresponding to a price-to-sales ratio of 240.
Kalshi was traded at a valuation of $11 billion and revenue of $264 million, approximately 42 times earnings.
DraftKings has a price-to-sales ratio of only 2.9.
Venture capital firms seem to be pouring money into these platforms almost nonstop, while at the same time, the stock prices of "traditional players" like DraftKings and Flutter Entertainment (whose subsidiary FanDuel) continue to be under pressure.
This once again confirms a familiar signal: capital is paying a premium for "potential future monetization" rather than paying for current profits.
Polymarket's valuation implies one of two possibilities: either it will unlock massive monetization potential in the future, or it will evolve into something far more than a "prediction market." At a revenue multiple of over 200, you're not buying a mature company, but rather a call option betting on a completely new financial primitive.
Perhaps Polymarket will become the default trading venue for hedging any real-world event; perhaps it will cover more sports, financial reports, weather, or any binary outcome event; perhaps it will raise its commission from 0.15% to a higher level, and its revenue will jump to billions of dollars overnight.
This is precisely the purest form of the "integration problem": In the future, will it belong to regulated exchanges with clear commission rates and compliance departments? Or will it belong to protocols that require no permission, allow anyone to place bets on any event anywhere, and where "the house takes almost no commission"?
The Convergence
A few years ago, we couldn't even put DeFi and Fintech on the same table for a direct comparison. Now, we're standing right there.
The crypto world has built a financial infrastructure that rivals fintech in terms of transaction volume, user base, and asset size: stablecoin channels are more globalized than traditional payment institutions; Aave's lending volume exceeds that of Klarna; and Polymarket's transaction volume is comparable to DraftKings.
The technology is feasible, and the product has found a sufficiently large user base. But the problem lies in value capture.
In every category we examined, the conclusions were highly consistent: crypto systems have lower commission rates and therefore capture less economic value compared to traditional fintech.
Encryption builds the most efficient and open infrastructure, at the cost of distributing value more widely.
Whether this is a bug or a feature depends on your perspective: if you believe that financial services will eventually evolve into commoditized public utilities, then crypto simply accelerates this inevitable process; if you believe that businesses must rely on revenue to survive, then most tokens still face serious challenges in value capture.
Regardless, convergence is already underway: banks are piloting tokenized deposits; the New York Stock Exchange is exploring tokenized stock trading; and the total supply of stablecoins has reached a new high of over $300 billion.
The existing giants in fintech have already recognized the trend—they won't ignore it, but rather absorb it.
The question for the next ten years is actually quite simple: will cryptography learn to build tollbooths, or will fintech learn to walk the crypto path?
Our assessment is that both will happen.
